OpenAI has identified a covert influence operation known as “Spamouflage,” which has been active on over 50 social media platforms. This operation primarily targeted users in the United States, Taiwan, the United Kingdom, Australia, and Japan, disseminating positive information about China. Initially identified by Meta in 2023, the operation involved using OpenAI’s ChatGPT to generate short-form material, including comments critical of Cai Xia, a Chinese political dissident.
Sophisticated Tactics and Targeting of Latin America
The operation demonstrated sophistication by prompting ChatGPT to generate Spanish-language articles during mainland Chinese working hours. The generated content was then posted on the platform X by users claiming to hail from the United States or India. OpenAI suspects these users leveraged ChatGPT to compose Spanish-language newspaper articles critical of the US, focusing on political divisions and societal issues such as drug use and homelessness. These articles found their way into mainstream media in Mexico, Peru, and Ecuador.
OpenAI has taken action against several accounts linked to China used to generate this propaganda material. Evidence suggests users employed ChatGPT to translate receipts from Latin American newspapers, implying that some articles might have been paid placements. The operation is particularly notable for its targeting of mainstream media outlets, a strategy OpenAI describes as unique.
“This is the first time we’ve observed a Chinese actor successfully planting long-form articles in mainstream media to target Latin America audiences with anti-US narratives, and the first time this company has appeared linked to deceptive social media activity.”
Originates from: OpenAI
In addition to the activities linked to China, OpenAI reported that groups based in Russia, Iran, and Israel have utilized its AI models in 2024 to generate short comments on social media. Among these groups is a Russian propaganda entity known as Bad Grammar, which used OpenAI’s technology to fabricate replies about Ukraine on Telegram.
The “Spamouflage” operation is characterized by its dual approach, combining traditional social media manipulation with targeted efforts in mainstream media. OpenAI believes this strategy may have reached a significantly broader audience than typical disinformation campaigns.
“a previously unreported line of effort, which ran in parallel to more typical social media activity, and may have reached a significantly wider audience.”
Originates from: OpenAI
Author’s Opinion
The discovery of the “Spamouflage” operation sheds light on the increasingly sophisticated methods used to manipulate public opinion globally. The blend of social media manipulation and integration into mainstream media demonstrates a new level of influence that AI-generated content can achieve. This tactic not only targets specific regions with anti-US narratives but also undermines the integrity of independent media outlets. As disinformation campaigns evolve, it is crucial for platforms, governments, and organizations to implement stronger safeguards to prevent such covert operations from affecting public discourse.
